Skip to content
Lacren

Private models. Paid access.

Keep the weights. Share the model.

Let people use your private AI model through an API. Set the price, keep control, and get paid when it runs.

See how protection works

Example classifierExample · illustrative
Price per successful request
$0.08/ request (illustrative)
Model weights
Weights stay private
Who can use it
Set by owner
Paid out
Per successful request

Example only. Not a live model, price or runtime check. Real terms and protection depend on each model and its verified deployment.

How it works

From private file to paid endpoint

Three steps, and the weights never leave your control.

  1. Upload privately

    Your model is encrypted on this device before upload.

  2. Choose the terms

    Choose who can use it and what each request costs.

  3. Get paid when it runs

    You earn for each successful request, settled on chain.

The same model. Different permissions.

Ownership stays on one side. Access crosses to the other, on the owner's terms.

OwnerExample
Weights
Not shared
Price
Set by you
Access
Public, unlisted or private
Earnings
Settled on chain
CustomerExample
Receives
API responses
Pays
Per successful request
Can inspect
Runtime proof and receipts
Keeps
Unused funds

Why private

Access crosses the boundary. Ownership doesn't.

  • The owner

    Keeps the weights. Only an attested runtime can unwrap the key.

  • The buyer

    Receives responses and a receipt for each charge.

  • The host

    Sees only what the documented security level allows.

Who can see what
ItemOwnerHostBuyer
Model weightsHas the fileEncrypted onlyNever
Customer inputsNot stored for themDepends on the modelSends them
ChargesSigned receiptsPublic on chainSigned receipts

For developers

One endpoint. Your model stays private.

One request, one exact response shape. Retries with the same idempotency key never charge twice.

Call a model
curl -X POST 'https://api.lacren.xyz/v1/inference/mdl_example' \
  -H "Authorization: Bearer $LACREN_API_KEY" \
  -H 'Idempotency-Key: your-unique-id' \
  -H 'Content-Type: application/json' \
  -d '{"inputs":{"features":[0.12,0.43,0.72]}}'
Response (shape only, illustrative values)
{
  "id": "call_01...",
  "model": "mdl_example",
  "version": "v1",
  "status": "completed",
  "output": { "score": 0.91 },
  "usage": { "unit": "request", "count": 1 },
  "billing": { "asset": "USDG", "amount": "0.08", "receiptId": "rcpt_..." },
  "protection": { "level": "weights-tee", "attestationId": "att_..." },
  "timing": { "totalMs": 0, "runtimeMs": 0 }
}

Pricing

Pay for what you run.

No subscriptions and no starting-at prices. Each model sets its own price per successful request.

For model owners

Pricing preview
8%proposed platform fee on paid requests

You set the price. Lacren keeps a share of each paid request and you keep the rest.

  • Hosting at the provider's live quote, shown with its expiry before you publish
  • The fee in effect is fixed in every access agreement
  • Withdraw settled earnings at any time

For model users

Per request

The model's price per successful request, plus network fees.

  • See the exact price before you connect a wallet
  • Failed requests have no per-request charge
  • Unused funds remain yours

See pricing details

Questions

Still unsure? The security page lists exactly what is and isn't covered.

Can buyers or Lacren access my model weights?

Buyers receive responses, never the weight file. The model is encrypted on your device before upload, and the key is released only to a runtime that passes a hardware check against an approved software measurement. Ordinary Lacren servers never hold the unwrapped key.

What does the runtime evidence show?

Each model has a proof page listing the machine check, the approved software check and the model-loaded check, each with a timestamp, plus the quote hash, verifier version and policy. If evidence is missing or expired, the model shows Not verified and stops taking new requests.

What happens to unused funds?

Funds you add stay in the Lacren escrow contract under your account. Anything not reserved for a request in progress can be withdrawn. Charges already settled still apply.

How fast are requests?

It depends on the model and runtime. Model pages show measured response times only once real data exists.

Why USDG?

Payments run on Robinhood Chain. Your wallet needs USDG and a little ETH for network fees.

Which hardware does protection depend on?

Protection relies on a confidential-computing CPU environment and its vendor's attestation. Hardware bugs and side channels remain possible, and anyone can learn about a model by repeatedly using its API. The security page lists what is and isn't covered.

Private models. Paid access.

Let people use your private AI model through an API. Set the price, keep control, and get paid when it runs.